单目运动目标的位姿实时测量方法
|
摘要
针对如何快速实时地获取运动目标位姿参数的问题,该文提出了一种单目运动目标位姿实时测量的方法,设计了标志点像物方匹配检测状态机和快速跟踪匹配状态机。匹配检测状态机在无标志点差异的条件下,采用反投影法得到最佳物像匹配对应关系,为快速高效跟踪匹配提供位姿初始解;提出了一种基于精度评定的状态切换机制,既发挥了跟踪匹配的高效性,又避免跟踪失败造成的位姿测量永久失败。实验结果表明,反投影法能够有效地筛选出最佳的像物方匹配对应关系,快速跟踪匹配状态机能够通过跟踪标志点高效地计算匹配关系和位姿参数,状态切换机制及时地判断跟踪错误,并迅速切换至匹配检测状态,实时测量全过程能高稳健运行。
3 Dposition and attitude estimation is used to get objects' position and attitude parameters in order to render the object in the computer.In order to obtain the pose parameters of the moving target in real time,this paper presents an approach for monocular real-time position and attitude estimation of moving objects,and introduces two mechanisms:detecting correspondence between LED(Light-Emitting Diode)markers on the target and their respective images,and fast tracking.On condition that there is no difference in shape of LED markers,the former mechanism uses back-projection method to find the best correspondence result,which provides the initial solution for fast-tracking mechanism.In addition,the paper also presents a status-switching mechanism based on accuracy assessment,which keeps the high efficiency of the fast-tracking matching and avoids permanent failure of position and attitude estimation due to the tracking failure.Extensive experiments illustrate that the back-projection method can filter out the best correspondence,and according to tracking markers,fast-tracking mechanism can calculate correspondence and pose parameters efficiently.The status-switching mechanism can identify tracking errors and switch to detecting mechanism quickly,in that the whole measure process is highly stable.
3 Dposition and attitude estimation is used to get objects' position and attitude parameters in order to render the object in the computer.In order to obtain the pose parameters of the moving target in real time,this paper presents an approach for monocular real-time position and attitude estimation of moving objects,and introduces two mechanisms:detecting correspondence between LED(Light-Emitting Diode)markers on the target and their respective images,and fast tracking.On condition that there is no difference in shape of LED markers,the former mechanism uses back-projection method to find the best correspondence result,which provides the initial solution for fast-tracking mechanism.In addition,the paper also presents a status-switching mechanism based on accuracy assessment,which keeps the high efficiency of the fast-tracking matching and avoids permanent failure of position and attitude estimation due to the tracking failure.Extensive experiments illustrate that the back-projection method can filter out the best correspondence,and according to tracking markers,fast-tracking mechanism can calculate correspondence and pose parameters efficiently.The status-switching mechanism can identify tracking errors and switch to detecting mechanism quickly,in that the whole measure process is highly stable.
引文
[1]吕博.新型空间飞行器交会对接光学系统研究[D].长春:中国科学院研究生院(长春光学精密机械与物理研究所),2015.(LYU Bo.Study on new optical systemof space craft rendezvous and docking[D].Changchun:Changchun Institute of Optics,Fine Mechanics andPhysics Chinese Academy of Sciences,China,2015.)
[2]徐剑.单通道旋转导弹制导系统一体化设计[D].上海:上海交通大学,2014.(XU Jian.Design of singlechannel rotary missile integrated guidance system[D].Shanghai:Shanghai Jiao Tong University,2014.)
[3]孙丽霞.高速列车横向运动稳定性和动态脱轨理论分析及评价方法研究[D].北京:中国铁道科学研究院,2014.(SUN Lixia.Evaluation method study and theoryanalysis of lateral running stability and dynamic derail-ment of high speed train[D].Beijing:China Academyof Railway Sciences,2014.)
[4]支春阳.GPS航姿仪的设计与实现[D].哈尔滨:哈尔滨工程大学,2010.(ZHI Chunyang.Design and imple-mentation of attitude measurement system based on GPS[D].Harbin:Harbin Engineering University,2010.)
[5]张开禾,富立,范耀祖.基于DSP和FPGA的捷联航姿仪的设计[J].测控技术,2007(3):75-76,78.(ZHANGKaihe,FU Li,FAN Yaozu.Design of strap-downattitude and heading reference system apparatus basedon DSP and FPGA[J].Measurement &ControlTechnology,2007(3):75-76,78.)
[6]ZHANG S J,CAO X B,ZHANG F,et al.Monocularvision-based iterative pose estimation algorithm fromcorresponding feature points[J].Science China(Information Sciences),2010,53(8):1682-1696.
[7]TERUI F.Model based visual relative motion estima-tion and control of a spacecraft utilizing computergraphics[C]∥Twenty-first International Symposiumon Spaceflight Dynamics(ISSFD).Toulouse,France:[s.n.],2009.
[8]KELSEY J M,BYRNE J,COSGROVE M,et al.Vision-based relative pose estimation for autonomousrendezvous and docking[C]∥Aerospace Conference.[S.l.]:IEEE,2006:20.
[9]GAO X,LIANG B,DU X,et al.Pose measurement oflarge non-cooperative satellite using structured lightvision sensor[C]∥IEEE International Conference onInformation Science and Technology.[S.l.]:IEEE,2012:101-108.
[10]SEGAL S,CARMI A,GURFIL P.Vision-basedrelative state estimation of non-cooperative spacecraftunder modeling uncertainty[C]∥In Aerospace Conference.[S.l.]:IEEE,2011.
[11]RUEL S,ENGLISH C,CHURCH P.3DLASSO:Realtime pose estimation from 3Ddata for autonomoussatellite servicing[C]∥I-sairas-international Symposiumon Artificial Intelligence.′i-SAIRAS 2005′-The 8thInternational Symposium on Artificial Intelligence,Robotics and Automation in Space,2005.
[12]XING Y,CAO X,ZHANG S,et al.Relative positionand attitude estimation for satellite formation withcoupled translational and rotational dynamics[J].ActaAstronautica,2010,67(3):455-467.
[13]PRESSIGOUT M,MARCHAND E,MEMIN E.Hybrid tracking approach using optical flow and poseestimation[C]∥IEEE International Conference onImage Processing.[S.l.]:IEEE,2008:2720-2723.
[14]IZADI S,KIM D,HILLIGES O,et al.KinectFusion:realtime 3Dreconstruction and interaction using a movingdepth camera[C]∥ACM Symposium on User InterfaceSoftware &Technology.[S.l.:s.n.],2011:559-568.
[15]LOWE D G.Distinctive image features from scale-invariant keypoints[C]∥International Journal ofComputer Vision.[S.l.]:[s.n.],2004:91-110.
[16]谭凯,张永军,童心,等.国产高分辨率遥感卫星影像自动云检测[J].测绘学报,2016(5):581-591.(TANKai,ZHANG Yongjun,TONG Xin,et al.Automaticcloud detection for Chinese high resolution remotesensing satellite imagery[J].Acta Geodaetica etCartographica Sinica,2016(5):581-591.)
[17]邓小炼,王星,艾思敏,等.奇异值分解和阈值分割的遥感自适应变化检测[J].测绘科学,2016,41(8):38-42.(DENG Xiaolian,WANG Xing,AI Simin,et al.Adaptive change detection algorithm of remote sensingimages based on SVD and OTSU[J].Science ofSurveying and Mapping,2016,41(8):38-42.)
[18]徐亚明,孙雪妍,王震,等.基于方向膨胀的核安全壳锈蚀识别算法[J].测绘通报,2017(3):13-16.(XUYaming,SUN Xueyan,WANG Zhen,et al.Rustingrecognition algorithm for the containment of the nuclearpower plant based on contour direct[J].Bulletin ofSurveying and Mapping,2017(3):13-16.)
[19]刘亦书,杨力华,孙倩.轮廓矩不变量及其在物体形状识别中的应用[J].中国图象图形学报,2004(3):54-59.(LIU Yishu,YANG Lihua,SUN Qian.Contour-based moment invariants and their application to therecognition of object shapes[J].Journal of Image andGraphics,2004(3):54-59.)
[20]邓非,吴幼丝,胡玉雷,等.球形全景影像位姿估计的改进EPnP算法[J].测绘学报,2016(6):677-684.(DENG Fei,WU Yousi,HU Yulei,et al.Position andpose estimation of spherical panoramic image withimproved EPnP algorithm[J].Acta Geodaetica etCartographica Sinica,2016(6):677-684.)
[2]徐剑.单通道旋转导弹制导系统一体化设计[D].上海:上海交通大学,2014.(XU Jian.Design of singlechannel rotary missile integrated guidance system[D].Shanghai:Shanghai Jiao Tong University,2014.)
[3]孙丽霞.高速列车横向运动稳定性和动态脱轨理论分析及评价方法研究[D].北京:中国铁道科学研究院,2014.(SUN Lixia.Evaluation method study and theoryanalysis of lateral running stability and dynamic derail-ment of high speed train[D].Beijing:China Academyof Railway Sciences,2014.)
[4]支春阳.GPS航姿仪的设计与实现[D].哈尔滨:哈尔滨工程大学,2010.(ZHI Chunyang.Design and imple-mentation of attitude measurement system based on GPS[D].Harbin:Harbin Engineering University,2010.)
[5]张开禾,富立,范耀祖.基于DSP和FPGA的捷联航姿仪的设计[J].测控技术,2007(3):75-76,78.(ZHANGKaihe,FU Li,FAN Yaozu.Design of strap-downattitude and heading reference system apparatus basedon DSP and FPGA[J].Measurement &ControlTechnology,2007(3):75-76,78.)
[6]ZHANG S J,CAO X B,ZHANG F,et al.Monocularvision-based iterative pose estimation algorithm fromcorresponding feature points[J].Science China(Information Sciences),2010,53(8):1682-1696.
[7]TERUI F.Model based visual relative motion estima-tion and control of a spacecraft utilizing computergraphics[C]∥Twenty-first International Symposiumon Spaceflight Dynamics(ISSFD).Toulouse,France:[s.n.],2009.
[8]KELSEY J M,BYRNE J,COSGROVE M,et al.Vision-based relative pose estimation for autonomousrendezvous and docking[C]∥Aerospace Conference.[S.l.]:IEEE,2006:20.
[9]GAO X,LIANG B,DU X,et al.Pose measurement oflarge non-cooperative satellite using structured lightvision sensor[C]∥IEEE International Conference onInformation Science and Technology.[S.l.]:IEEE,2012:101-108.
[10]SEGAL S,CARMI A,GURFIL P.Vision-basedrelative state estimation of non-cooperative spacecraftunder modeling uncertainty[C]∥In Aerospace Conference.[S.l.]:IEEE,2011.
[11]RUEL S,ENGLISH C,CHURCH P.3DLASSO:Realtime pose estimation from 3Ddata for autonomoussatellite servicing[C]∥I-sairas-international Symposiumon Artificial Intelligence.′i-SAIRAS 2005′-The 8thInternational Symposium on Artificial Intelligence,Robotics and Automation in Space,2005.
[12]XING Y,CAO X,ZHANG S,et al.Relative positionand attitude estimation for satellite formation withcoupled translational and rotational dynamics[J].ActaAstronautica,2010,67(3):455-467.
[13]PRESSIGOUT M,MARCHAND E,MEMIN E.Hybrid tracking approach using optical flow and poseestimation[C]∥IEEE International Conference onImage Processing.[S.l.]:IEEE,2008:2720-2723.
[14]IZADI S,KIM D,HILLIGES O,et al.KinectFusion:realtime 3Dreconstruction and interaction using a movingdepth camera[C]∥ACM Symposium on User InterfaceSoftware &Technology.[S.l.:s.n.],2011:559-568.
[15]LOWE D G.Distinctive image features from scale-invariant keypoints[C]∥International Journal ofComputer Vision.[S.l.]:[s.n.],2004:91-110.
[16]谭凯,张永军,童心,等.国产高分辨率遥感卫星影像自动云检测[J].测绘学报,2016(5):581-591.(TANKai,ZHANG Yongjun,TONG Xin,et al.Automaticcloud detection for Chinese high resolution remotesensing satellite imagery[J].Acta Geodaetica etCartographica Sinica,2016(5):581-591.)
[17]邓小炼,王星,艾思敏,等.奇异值分解和阈值分割的遥感自适应变化检测[J].测绘科学,2016,41(8):38-42.(DENG Xiaolian,WANG Xing,AI Simin,et al.Adaptive change detection algorithm of remote sensingimages based on SVD and OTSU[J].Science ofSurveying and Mapping,2016,41(8):38-42.)
[18]徐亚明,孙雪妍,王震,等.基于方向膨胀的核安全壳锈蚀识别算法[J].测绘通报,2017(3):13-16.(XUYaming,SUN Xueyan,WANG Zhen,et al.Rustingrecognition algorithm for the containment of the nuclearpower plant based on contour direct[J].Bulletin ofSurveying and Mapping,2017(3):13-16.)
[19]刘亦书,杨力华,孙倩.轮廓矩不变量及其在物体形状识别中的应用[J].中国图象图形学报,2004(3):54-59.(LIU Yishu,YANG Lihua,SUN Qian.Contour-based moment invariants and their application to therecognition of object shapes[J].Journal of Image andGraphics,2004(3):54-59.)
[20]邓非,吴幼丝,胡玉雷,等.球形全景影像位姿估计的改进EPnP算法[J].测绘学报,2016(6):677-684.(DENG Fei,WU Yousi,HU Yulei,et al.Position andpose estimation of spherical panoramic image withimproved EPnP algorithm[J].Acta Geodaetica etCartographica Sinica,2016(6):677-684.)